On 18 May 2012, at 18:45, Erick Britis Ortiz <[log in to unmask]> wrote:
> Dear Vladimir,
> tested the fix with several projects, it seems to work fine.
> One related
> question: spm_cond_units.m was changed, which is also used by
> spm_robust_average.m and spm_robust_glm.m. Could this create problems?
No, because those functions store the scaling factor and reapply it at the end to restore the original scaling.
> And, not unrelated, but something that has been long on my mind:
> If I want to make a group analysis, bringing the final images as
> 1st-level results, can I use use them in the same way I would use the
> beta images from fMRI? Do these intermediate re-scalings, and different
> temporal modes across subjects induced different scales in the image values?
> I used to think that group inversion was necessary, else they would have
> different spatial modes. But how can statistics even be done in a group,
> if the units are lost and "effect size" does not have the same meaning
> as in fMRI?
You can do group statistics with or without group inversion but all the conditions and time windows within subject that you want to compare should be inverted together and the images written out together. The scaling is applied to all conditions jointly so their relations should be preserved.
> On 2012-05-08 23:51, Vladimir Litvak wrote:
>> Dear Stephan and Erick,
>> Please try the attached fix for the problems with lead field scaling.
>> It does seem to improve things on our example. It'll be in the next
>> SPM8 update.